Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Nyquist double sample rate rule (Read 13388 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Nyquist double sample rate rule

Reply #25
@bennetng It is unclear what are you trying to add to a two year old thread.
Well, because it is still on the first page of this subforum.
btw the question was 'what', not 'why'

I was trying to figure out the accuracy issue / limitation is on the waveform display or the waveform generator.
it's the waveform display.

I can't really see things clearly.
you can try pulse train instead of white noise (for example, from http://src.infinitewave.ca/TestSignals.zip )

Re: Nyquist double sample rate rule

Reply #26
The best looking Audition screenshot setting I can get when using SSRC to upsample the 96k pulse train to 192k:
X

Re: Nyquist double sample rate rule

Reply #27
This ripple in the passband looks strange.

Re: Nyquist double sample rate rule

Reply #28
With the Audition FFT display I can only get a flat passband when analyzing a single pulse, but the amplitude will be too low in a way that the transition region will fall below the lowest allowed dB display range (-220dB).

I could use other software like Deltawave of course, but just to show that the Audition FFT display has limited use for these analyses. Also, the waveform view indeed has some interpolation when zoomed in, but not high precision enough to avoid visual artifact when the frequency is too close to Nyquist.

[EDIT]How to do it with a single pulse on Audition, highly frustrating though. Select a range of samples longer than the FFT size, click "scan", manually set the reference dB level to prevent it going down too low. The FFT noise floor is not limited by the file format, but the FFT display, which is a trap. Read this post for an example:
https://hydrogenaud.io/index.php?topic=118758.0
X

Re: Nyquist double sample rate rule

Reply #29
You can try sox's upsample, sinc, gain:
Actually "vol 6" instead of "gain" would be better in this situation.

Thanks. I am too lazy but is it steeper than rate -v or -u at 99.7?
Looks like it
Code: [Select]
sox noise.8k.wav -r48k noise.48k.sinc.wav upsample 6 sinc -3993 -t 1 vol 6
sox noise.8k.wav noise.48k.rate-vb.wav rate -v -b 99.7 48k
sox noise.8k.wav noise.48k.rate-ub.wav rate -u -b 99.7 48k
X


Re: Nyquist double sample rate rule

Reply #31
Looks like it
Ok, I've just found out the comparison wasn't too fair. The "sinc" default attenuation level is 120 dB and rate's -v/-u 175 dB. I re-did it in 32 bits and compared:
  • sinc -a 125 vs rate -h
  • sinc -a 175 vs rate -v vs rate -u
X
X

Re: Nyquist double sample rate rule

Reply #32
Top: 20k sine at 44.1k sample rate like what the Monty video shows.
Bottom: 21k.
Audition's precision could be similar to what typical real world hardware actually do. Many DACs use half band filters too, with stopband at about 24kHz when running at 44.1k.
X

Re: Nyquist double sample rate rule

Reply #33
Looks to me like perhaps the image at FS - f isn't getting adequately filtered so you're just seeing (and hearing at lower frequencies) the beating between the 2 tones.

Re: Nyquist double sample rate rule

Reply #34
Top: 20k sine at 44.1k sample rate like what the Monty video shows.
Bottom: 21k.
Audition's precision could be similar to what typical real world hardware actually do. Many DACs use half band filters too, with stopband at about 24kHz when running at 44.1k.
Looks similar as in sonic-visualizer:
X

sonic-visualizer is using rather gentle filter for 8x oversampling (assuming I properly converted its coefficients to spectrum):
X

 

Re: Nyquist double sample rate rule

Reply #35
My Creative soundcard's DAC (TI PCM1794A) at 44.1k recorded by Realtek ALC897's ADC at 192kHz, which also shows stereo crosstalk and some jitter sidebands, and reduced amplitude of the 21kHz tone due to near Nyquist filtering.
X